Generative Engine Optimization SaaS
TopoRAG
A Chrome Extension SaaS that applies topological data analysis to LLM embeddings, ranking semantic content gaps competitors leave open.
TDA + pgvector + production billingAI Engineer + Product Builder
I'm Mustapha Liaichi, an AI engineer at AYAutomate. I ship LLM products, agent infrastructure, automation systems, and the occasional unusual experiment that turns into a useful tool.

Based in Rabat, Morocco
Building for the open web
Building
Independent products, browser tools, and the studio behind them.
Developer storytelling
Turn GitHub repositories into project stories, with the work behind each build ready to share.
02 / BuildingChrome extension
Find people on Reddit who need what you sell, and collect relevant conversations in a side-panel inbox.
Apify Actor builder
Describe the data you need. Build, test, and publish an Apify Actor with source code you can inspect and keep.
AI engineering studio
Build AI products, agents, and automation around real workflows, with the code and a clear handover.
Selected systems
A cross-section of work spanning SaaS products, agent memory, marketplace intelligence, fine-tuning, and open protocols.
Generative Engine Optimization SaaS
A Chrome Extension SaaS that applies topological data analysis to LLM embeddings, ranking semantic content gaps competitors leave open.
TDA + pgvector + production billingCognitive database for AI agents
An AI-native memory engine combining episodic, semantic, and procedural stores with token-aware retrieval, decay, and consolidation.
Adapters for AutoGen, LangGraph, and MCPSales intelligence for automation freelancers
Pre-meeting company briefings that surface automation opportunities, pain points, discovery questions, and ready-to-send outreach.
Native MCP server with persistent API keysMarketplace intelligence actor
Analyzes public Apify Store data to uncover demand, pricing patterns, category momentum, competitors, and promising actor ideas.
Built from weekly analysis of 10,000+ actorsFine-tuned Llama 3 model
A production model fine-tuned on thousands of real automation examples to generate structured n8n workflows from natural language.
8B parameters, Apache 2.0Federated Agent Protocol
The first governed LLM-backed participant in an open federated coordination protocol, extended for Arabic, Darija, and French agents.
Open source, multilingual agent coordinationBuilder loop
My public work connects product development with market research and education. I analyze automation ecosystems, turn the findings into tools, and publish the practical lessons for other builders.
Updated
Research notes
Long-form explorations of LLM infrastructure, multi-agent systems, memory, AI engineering, and Morocco's emerging AI ecosystem.
Browse all 11 notesI hadn't forgotten how to code. I'd lost the map of my own projects. This is how I took back the parts I had handed over to AI.
Two papers published in 1991 disagreed on almost everything. The reactive versus deliberative debate set the terms for 35 years of multi-agent research.
A budget of 500 tokens with the right memories can beat 4,000 with random ones. This is the architecture behind CogDB.
Morocco is building an AI ecosystem from scratch. But sovereignty requires systems that work for the languages people actually speak.
Field notes
Market signals, automation patterns, experiments, and honest post-mortems.
Experience
AYAutomate
Designing end-to-end AI automation for international clients, from RAG and multi-agent workflows to intelligent document processing and reusable agent architecture.
Lear Corporation
Deployed Palantir Foundry AI workflows for automotive requirements, reaching 89% accuracy and cutting manual review time by 75%.
Sweet Spot AI
Led AI development for a social intelligence platform processing millions of interactions, with RAG, GPT-4 reporting, OAuth, and billing.
Independent
Delivered 50+ AI projects for international clients and built data pipelines handling more than 10 million records.
Certifications
A record of the skills I keep sharpening.

Anthropic
Building and integrating applications and agents with the Claude API, Claude Code, custom tools, and MCP servers.
Issued to Mustapha LiaichiWorking set
Engineering degree in Data Science and Knowledge Engineering from the School of Information Sciences in Rabat.
LLMs, RAG, multi-agent architecture, TDA on embeddings, prompt engineering
MCP, LangChain, CrewAI, Mastra, AutoGen, LangGraph, Hugging Face
Python, TypeScript, FastAPI, Next.js, Django, PostgreSQL, Supabase
Palantir Foundry, Vercel, Railway, Docker, AWS, Azure, pgvector